Research And Implementation Of Face Matching Algorithm Based On Gabor And Depth Learning

Face recognition is not only the basis of computer vision field, but also the core of visual field. Image classification and people's lives are closely related. In recent years, due to the depth of learning methods in ImageNet ILSVRC contest brilliant achievements, image classification research is increasingly active. The advent of the Big Data era provides unprecedented opportunities for the development of artificial intelligence.In context breakthroughs in depth learning, including image classification,are not accidental.The traditional methods of face recognition are very limited, such as Gabor and machine learning. It is not the first time to apply depth learning to face recognition. For the first time, CNN has proposed a deep learning-based image recognition process, which extracts features from candidate regions by using deep convolution network. Then, it classifies regions into objects and backgrounds based on features using linear classifiers, such as support vector machines. In this paper, an improved face recognition system based on depth learning is implemented by improving CNN model. In this paper, we combine face recognition with Gabor and CNN. Firstly, Gabor filtering is applied to face image.Secondly, the Gabor processed images are input into the network to reduce the computations, and then CNN is modified to modify the CNN network structure and improve the average accuracy rate (mAP) of MultiPIE datasets. In addition, our improved face recognition algorithm based on CNN greatly reduces the computational cost of the network parameters and improves the speed to meet the real-time requirements.Finally, our face classification algorithm in MultiPIEdata set obtained 91% accuracy, compared to simple Gabor filter model increased by 6%,compared to the traditional CNN increased by 3%.